Thread by @QuinnyPig

The peoples have spoken, so it's time to tweet about that time I used Route53 as a database. A thread of misery...

So once upon a time, many years ago, I was a grumpy UNIX systems administrator--or "a sysadmin," as there really aren't any other kinds.

We ran some EC2 instances that themselves ran "containers," (OpenVZ) because this was many years ago and everything was terrible.

We had a problem. Thanks to a variety of poor decisions, there was no easy way to figure out what container lived on top of which instance. These were long lived; it mattered.

The exact poor decision tree isn't the fun part of the story, so I'll handwave past that. Please enjoy this visual metaphor instead:
🦘💨💩

So we now have an internal DNS zone where we can get into all of our various containers. They were named reasonably, as it was two years past my "characters from the Dune novels" computer naming phase.

...which was itself a year or two past my "Federation starships" naming phase. I digress.

So we'd have a naming convention in DNS such as web003.prod.iad.twitterforpets.com or similar. You could target that with a regular expression, you could use globbing for bulk actions, you could screw up a wildcard and destroy your career, etc.

I'm not done.

You could *also* figure out which node that sucker lived on with a `dig TXT web003.prod.iad.twitterforpets.com` and get back `…rvisor001.prod.iad.twitterforpets.com`.

"Well, they're called text records, may as well use them to store useful info!"

While it's far from the dumbest thing I've ever done with DNS, somewhere @dbsmasher just sat bolt upright, heart racing. She knows not why until she checks Twitter.

And that, folks, is why I don't ever accept "hands on keyboard" projects with my clients. Because I will build monstrosities like this, and somehow make it sound like a good idea.